Predictive Models for Estimation of Swelling Characteristics of Expansive Soils Based on the Index Properties

The soils that exhibit volume changes with change in moisture content are called expansive or swelling soils. These soils are characterized generally by their blackish colour, high plasticity and the enriched presence of the clay mineral montmorillonite as the principal constituent. As the expansive soils have a tendency to change its volume with change in the moisture content, they can cause severe damage and distress to lightweight structures constructed over them due to the increased swell pressure evolved as a result of the swelling. Hence while designing the foundations on expansive soils, it is highly imperative to get an idea of the anticipated swell and the associated swell pressure that may damage the structural element. The various soil properties which clearly indicate the swelling characteristics of the expansive soils are free swell index, swelling index, shrinkage limit, swelling potential, swelling pressure, etc. The activity of clay derived from plasticity index and percentage of clay sizes present in the soil is also used as an indicator for identifying the expansive soil. If these swelling parameters can be derived from easily determinable index properties of soil, that would be really informative regarding the usefulness of the soil or the quantum of modification or improvement required for reclamation of such type of soil. A great deal of research has been done in correlating the swelling characteristics with the index properties and physical state of soil. The present work is an attempt to develop predictive models for the swelling characteristics based on their interrelationship with the index properties like liquid limit, plasticity index, shrinkage index, etc. and the physical properties like dry unit weight, grain size, especially the clay percentage, etc. The expansive soils investigated at Central Soil and Materials Research Station, New Delhi, are used to develop the predictive models. The applicability of the developed equations is finally checked by conducting validation study using three different data sets.